Robots can already sprint. The problem is getting them to think.
The hardware in modern robots is impressive, but the software that makes them useful for real, general work is still years behind — and the companies betting on closing that gap are raising huge sums right now.
- $670M+
- raised in two years
- $2.9B
- current valuation
- 90%
- task success rate in structured settings
- 8+ years
- until home deployment
Robots have become impressive very fast. Chinese humanoids can sprint, dance, and do backflips. But the software inside them is still years behind the hardware — and that gap is where the real opportunity lives.
Spirit AI, a two-year-old Chinese startup, has already deployed its Moz1 wheeled humanoid robots on production lines at CATL, a giant battery maker, and JD.com, China's big retail platform. Its co-founder, Gao Yang, says a robot-brain breakthrough is coming by mid-2027 — something like what ChatGPT did for text AI in 2022, when a hard technology suddenly became something anyone could use.
Today, even the best industrial robots only succeed about 90% of the time on simple, structured tasks. Fine motor work — like unscrewing a bottle cap — still trips them up. Gao Yang calls the robot brain "the weakest link." To fix it, Spirit AI employs about 1,000 people across China doing everyday physical tasks so the AI can learn from messy, real-world motion data instead of tidy simulations.
Spirit AI has raised over $670 million in just two years and is now valued at $2.9 billion. Its robots work well in predictable, structured settings — factories and warehouses. Home robots are still 8 or more years away. The near-term opportunity is industrial, not domestic.
If you build products for manufacturing, logistics, or retail, the question worth asking now is not "will robots take over?" but "which specific workflows are ready, and in how long?" The window is opening. Teams that start mapping their most repetitive physical tasks today will be ready when the software catches up.